Overview

Fundly is a platform that helps small businesses acquire loans without speaking to a human agent. It uses an AI voice agent powered by ElevenLabs to fill out application fields through natural conversation, or accepts traditional text input for users who prefer typing. The system generates comprehensive loan evaluations with visual grade displays and personalized recommendations.

Problem

The traditional small business loan application process is inefficient and costly for both applicants and lending institutions. Businesses must schedule calls with loan officers, spending significant time explaining their situation. Lending institutions must hire staff to conduct these calls, manually process applications, and make loan decisions. This creates bottlenecks and limits the number of applications that can be processed, while increasing operational costs for lenders and creating delays for businesses that need quick access to capital.

Solution

An automated loan application platform that eliminates the need for human agents in the initial application phase:

Key Features:

  • AI voice assistant that conducts natural conversations to gather application data
  • Optional text input form for users who prefer typing over voice interaction
  • Comprehensive loan evaluation analyzing credit score, DTI ratio, LTV ratio, employment status, and loan type
  • Visual grade display (A+ to F) with detailed score breakdown for transparency
  • Automated transcript saving to database for loan officer review
  • AI-powered extraction of structured loan data from natural conversations using Claude AI
  • Personalized recommendations based on applicant profile and financial situation

Business Impact:

The platform eliminates scheduling conflicts, reduces staffing costs for lending institutions, and provides instant application processing for businesses. Loan officers can focus on reviewing structured data and making final decisions rather than conducting repetitive initial interviews.

Technical Architecture

Built on a modern web stack optimized for conversational AI and real-time data processing:

Frontend:

  • TypeScript for type-safe development
  • Tailwind CSS for responsive, modern styling

Backend & AI:

  • Python for AI integration and data processing
  • Supabase for scalable backend storage of transcripts and application data
  • ElevenLabs ConvAI for natural voice agent interactions
  • Claude AI for intelligent extraction of structured data from unstructured conversations
  • Automated loan scoring algorithm analyzing multiple financial factors

Data Processing:

The system processes voice conversations in real-time, extracts structured financial data, and generates comprehensive loan evaluations with grade displays. All transcripts are automatically saved for compliance and loan officer review.

Challenges

Creating a voice AI that could conduct natural financial conversations while accurately extracting structured data required sophisticated prompt engineering and conversation design. The system needed to handle diverse business scenarios and financial situations while maintaining data accuracy. Ensuring compliance with lending regulations while automating the process required careful system design and data validation. The platform needed to collect all necessary information for regulatory compliance without overwhelming applicants with technical jargon. Balancing automation with human oversight was crucial - the system needed to provide loan officers with structured, actionable data while flagging applications that required additional review or clarification.

Results & Impact

Fundly successfully automates the initial loan application process, reducing processing time from days to minutes while maintaining high standards for data accuracy and completeness. The platform provides instant feedback to applicants through visual grade displays and personalized recommendations. Lending institutions benefit from reduced operational costs and increased application throughput, while businesses gain immediate access to loan evaluation without scheduling delays. Loan officers can focus on decision-making rather than data collection, improving overall efficiency. The project demonstrates that AI can make financial services more accessible and efficient without compromising regulatory compliance or lending standards.